Variable selection for high-dimensional partly linear additive Cox model with application to Alzheimer's disease

Area of Science:

• Biostatistics
• Statistical Modeling
• Survival Analysis

Background:

• Extensive research exists for right-censored failure time data.
• Analysis of interval-censored data, especially with mixed-dimensionality covariates, presents unique challenges.

Purpose of the Study:

• To develop a penalized estimation procedure for simultaneous variable selection and estimation in interval-censored failure time data.
• To address situations with both low-dimensional (potentially nonlinear) and high-dimensional covariates.

Main Methods:

• Utilized Bernstein polynomials to approximate nonlinear covariate effects.
• Developed a coordinate-wise optimization algorithm for efficient implementation.
• Applied penalized estimation for simultaneous variable selection and estimation.

Main Results:

• The proposed method demonstrates effective simultaneous variable selection and estimation.
• Numerical studies indicate the approach performs well in practical scenarios.
• The method was successfully applied to an Alzheimer's disease study.

Conclusions:

• The penalized estimation procedure offers a robust solution for interval-censored failure time data analysis.
• The method is particularly valuable for complex datasets involving mixed-dimensional and nonlinear covariates.
• This work provides a practical tool for biomedical research, exemplified by its application in Alzheimer's disease.